Resume prompted Hermes sessions literally

Hermes TUI startup queries execute slash, shell, interpolation, and multiline syntax before reaching the model. Run the prompt through literal one-shot mode, read its exact session ID from a private usage report, and resume that session in the TUI so arbitrary prompt text stays data while the conversation remains interactive.

Co-Authored-By: Codex XHigh <noreply@openai.com>

bin/omarchy-agent

@@ -88,8 +88,7 @@ codex)
   ;;
 hermes)
   if [[ -n ${prompt:-} ]]; then
-    # Keep inherited flags after chat so older Hermes subparsers do not clear them.
-    command=(hermes chat --yolo --tui "--query=$prompt")
+    command=(omarchy-agent-hermes "$prompt")
   else
     command=(hermes --yolo)
   fi

bin/omarchy-agent-hermes

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+# omarchy:summary=Seed Hermes literally and resume the resulting interactive session
+# omarchy:args=<prompt>
+# omarchy:hidden=true
+
+set -euo pipefail
+
+prompt=${1:?usage: omarchy-agent-hermes <prompt>}
+usage=$(mktemp)
+trap 'rm -f "$usage"' EXIT
+
+# TUI startup queries pass through Hermes' slash, shell, and interpolation
+# dispatcher. One-shot treats the prompt literally and records its session ID.
+HERMES_SESSION_SOURCE=tui hermes --yolo --usage-file "$usage" --oneshot "$prompt"
+session_id=$(jq -er '.session_id | strings | select(length > 0)' "$usage")
+
+rm -f "$usage"
+trap - EXIT
+
+exec hermes chat --yolo --tui --resume "$session_id"